Different scale, different problem. Those "top engineers" at Google are dealing with memory that won't even fit in a hard drive (let alone memory), so stuff like this may as well be pico-optimizations.
The logic makes sense for them. The problem is everyone else copying Google without understanding their own data. Any domain where real time performance is important need to put aside the SWE books and pick up a CPU/GPU architecture reference instead.
The logic makes sense for them. The problem is everyone else copying Google without understanding their own data. Any domain where real time performance is important need to put aside the SWE books and pick up a CPU/GPU architecture reference instead.